| Metric | lstm | xgboost | naive | cnn |
|---|---|---|---|---|
| diff_of_means | 1.026 | 1.534 | 1.575 | -1.621 |
| ratio_of_sd | 0.946 | 0.909 | 0.914 | 0.949 |
| amplitude_ratio_of_means | 0.768 | 0.609 | 0.641 | 0.740 |
| maximum_error | 0.190 | 0.154 | 0.173 | 0.138 |
| ks_mean_on_coarse_res_with_extremes | 0.186 | 0.347 | 0.323 | 0.259 |
| qqplot_mae | 0.056 | 0.093 | 0.103 | 0.087 |
| acf_mae | 0.087 | 0.140 | 0.126 | 0.100 |
| extremogram_mae | 0.033 | 0.074 | 0.065 | 0.054 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | lstm | naive | xgboost | cnn |
|---|---|---|---|---|
| diff_of_means | 0.525 | 0.718 | 0.758 | -2.057 |
| ratio_of_sd | 0.952 | 0.909 | 0.899 | 0.957 |
| amplitude_ratio_of_means | 0.772 | 0.641 | 0.609 | 0.748 |
| maximum_error | 0.198 | 0.174 | 0.152 | 0.145 |
| ks_mean_on_coarse_res_with_extremes | 0.142 | 0.245 | 0.324 | 0.194 |
| qqplot_mae | 0.057 | 0.115 | 0.106 | 0.105 |
| acf_mae | 0.078 | 0.118 | 0.130 | 0.092 |
| extremogram_mae | 0.013 | 0.044 | 0.041 | 0.023 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | lstm | cnn | xgboost | naive |
|---|---|---|---|---|
| diff_of_means | 0.972 | -1.559 | 1.649 | 1.685 |
| ratio_of_sd | 0.952 | 0.960 | 0.891 | 0.897 |
| amplitude_ratio_of_means | 0.783 | 0.763 | 0.610 | 0.643 |
| maximum_error | 0.201 | 0.161 | 0.146 | 0.163 |
| ks_mean_on_coarse_res_with_extremes | 0.130 | 0.147 | 0.310 | 0.279 |
| qqplot_mae | 0.052 | 0.084 | 0.107 | 0.118 |
| acf_mae | 0.068 | 0.082 | 0.122 | 0.111 |
| extremogram_mae | 0.038 | 0.045 | 0.071 | 0.066 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | lstm | cnn | xgboost | naive |
|---|---|---|---|---|
| diff_of_means | 0.925 | -1.538 | -4.327 | -4.509 |
| ratio_of_sd | 0.954 | 0.968 | 0.971 | 0.990 |
| amplitude_ratio_of_means | 0.766 | 0.743 | 0.638 | 0.665 |
| maximum_error | 0.195 | 0.137 | 0.180 | 0.135 |
| ks_mean_on_coarse_res_with_extremes | 0.257 | 0.273 | 0.405 | 0.388 |
| qqplot_mae | 0.101 | 0.080 | 0.173 | 0.178 |
| acf_mae | 0.085 | 0.098 | 0.132 | 0.128 |
| extremogram_mae | 0.094 | 0.098 | 0.135 | 0.131 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | lstm | cnn | xgboost | naive |
|---|---|---|---|---|
| diff_of_means | 0.808 | -1.506 | 36.022 | 36.479 |
| ratio_of_sd | 0.959 | 0.970 | 0.546 | 0.610 |
| amplitude_ratio_of_means | 0.764 | 0.750 | 0.368 | 0.522 |
| maximum_error | 0.202 | 0.163 | 0.269 | 0.239 |
| ks_mean_on_coarse_res_with_extremes | 0.194 | 0.224 | 0.373 | 0.233 |
| qqplot_mae | 0.065 | 0.066 | 1.425 | 1.442 |
| acf_mae | 0.083 | 0.097 | 0.128 | 0.084 |
| extremogram_mae | 0.030 | 0.036 | 0.045 | 0.026 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | lstm | cnn | xgboost | naive |
|---|---|---|---|---|
| diff_of_means | 0.909 | -1.383 | 36.673 | 37.285 |
| ratio_of_sd | 0.949 | 0.960 | 0.529 | 0.599 |
| amplitude_ratio_of_means | 0.769 | 0.756 | 0.361 | 0.526 |
| maximum_error | 0.183 | 0.156 | 0.273 | 0.213 |
| ks_mean_on_coarse_res_with_extremes | 0.144 | 0.183 | 0.323 | 0.204 |
| qqplot_mae | 0.057 | 0.073 | 1.451 | 1.474 |
| acf_mae | 0.065 | 0.079 | 0.115 | 0.067 |
| extremogram_mae | 0.011 | 0.017 | 0.032 | 0.019 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | lstm | cnn | xgboost | naive |
|---|---|---|---|---|
| diff_of_means | 0.433 | -1.964 | 35.896 | 36.372 |
| ratio_of_sd | 0.948 | 0.954 | 0.543 | 0.607 |
| amplitude_ratio_of_means | 0.767 | 0.748 | 0.371 | 0.525 |
| maximum_error | 0.205 | 0.151 | 0.278 | 0.227 |
| ks_mean_on_coarse_res_with_extremes | 0.170 | 0.189 | 0.357 | 0.221 |
| qqplot_mae | 0.061 | 0.090 | 1.420 | 1.438 |
| acf_mae | 0.078 | 0.094 | 0.125 | 0.084 |
| extremogram_mae | 0.014 | 0.014 | 0.027 | 0.027 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97